Four gates 2: interrupts on, then the encoder

Gate 1 opened at 507 ms. At 512 ms BspInit switches on the 1 ms interrupt (TIM13) and the current ADC interrupt, but not the position timer TIM5. The current interrupt fires every 50 µs; its entry, CURRENT_LOOP_IRQ_TASK, checks gates 2 and 3 and skips the current loop while they are shut. On its very first run it does one more job: it starts TIM5, so both loops count the same 240 MHz clock with a fixed phase (two runners off one starting gun). Gate 2, hardware_self_test_status: the position interrupt checks the self-test before it reads the encoder reply, so the check only sees the encoder on tick 3 (512.25 ms), like a letter posted Monday, delivered Tuesday, read Wednesday. Encoder unplugged: it gives up after 100 ms, sets an error and retries every 100 ms; plugged in at 700 ms: the gate opens, but the error stays and the motor never moves. All from real runs of the board build.